HN Explorer

Ask HN: Deepfakes and publishing sha256sums of historical videos to a blockchain

4 points by orblivion ↗ HN
It's X years in the future. We can create deep fakes that can't be distinguished from the Real Thing. Especially for videos of older quality. We have a corpus of historical videos. We temporarily live under an oppressive regime that rewrites history to fit its needs. The institutions that we trust to attest to their authenticity (archive.org, C-SPAN, many others) have thus become compromised. We could lose the thread of certain historical documents _permanently_.

But somewhere, in the shadows, or in a free land somewhere in the world, a proof-of-work blockchain lives on. It contains timestamped sha256sums of manifest files of videos, which in turn contain sha256sums of the videos. It proves that certain pieces of data existed at certain times. Once the regime falls, the authentic videos can be determined by the hashes that were published before the regime came to power.

Is anybody talking about this scenario and/or remedy? Maybe there's a less alarmist version to consider as well.

8 comments

[ 2.6 ms ] story [ 32.6 ms ] thread
davidlee1435
I wish this were a thing, I really do. At the very least, I want to live in a future where we know if a piece of content was generated by a generative model or if it wasn't.

The question is- what is the incentive to burn gas (re: money) to timestamp the manifests before the regime comes into power?

ChainOfFools
a number of projects aimed at this exact problem - one of them is/was called Tyrion - there were probably a dozen because the concept is quite obvious. but the usual blockchain problems frustrate its success, namely that blockchains are not independent of human control, but only obfuscate it behind unprovable promises of decentralization.
orblivion
I would think that's only a forward-facing problem. I.e. policies could change. But so long as the chain has not been corrupted, and assuming it was proof-of-work, couldn't you more or less prove that the chain going backwards has in fact not been corrupted? Though I suppose it might rely on a reliable account of the hashing power of the world at a given time.
orblivion
These organizations are already incentivized, somehow or other, to do the work of creating and maintaining their archives. I think their incentives are aligned with wanting them to be authenticated. If you think gas fees would be a concern, they could create and publish a manifest file once a year for any new videos, and write the hash of the manifest file to the blockchain. That's one hash published a year, that can't be a lot.
orblivion
Though I guess if you're worried about whether something happened last week, you'll want more speedy publishing of hashes. But for any events that happen _in_ that future, we probably won't have much reason to trust videos anyway. That's a whole other problem. Fake news vs fake history.
vhcr
It doesn't has to be expensive, the files could be chunked in 1-day chunks for example, and a hash of all of them computed with a merkle tree.
lcnPylGDnU4H9OF
> Maybe there's a less alarmist version to consider as well.

I'll take a crack at this since it's ultimately where I land. People will just get wise to it and "videos" of stuff will eventually be seen as trustworthy as "someone's words" of stuff. There was a time before video technology and I suspect we're headed for something similar to that. At least with respect to trust in historical documentation.

hayst4ck
Trust is built on top of institutions. Focusing on strengthening institutions seems more useful than accurate engravings on their grave stone.

The powerful will never willingly give away their power. Developing tools that can take power away from the powerful (which will require a fight and personal risk for the fighters) or helping the powerless become more powerful seems infinitely more useful and practical.

"God made Man, Sam Colt made 'em equal..." Figuring out ways to make billionaires "equal" will do far more to help freedom and reduce oppression than some notion of historical book keeping. Figuring out how to credibly threaten consequences for billionaires seems like a much higher priority than helping future historians. Maybe the best threat is collective bargaining and tools to help collectively bargain would be a much better investment.

If the educated are able to exercise power, then the need for protecting history from those who wish to manipulate it isn't as much of a problem.